A lightweight Python framework for building multi-agent workflows with tools, handoffs, guardrails, sessions and tracing, provider-agnostic across OpenAI and 100+ LLMs.
vercel/eve
A filesystem-first TypeScript framework for building durable AI agents, with tools, skills, channels and schedules defined as files.
- Stars
- 5,641
- Forks
- 626
- Open issues
- 836
- Last commit
- 3 hours ago
What is it?
eve is a framework for building AI agents where the project structure is the authoring interface. An agent lives in an agent/ directory: instructions.md holds the always-on system prompt, while tools/, skills/, channels/ and schedules/ hold typed tool functions, on-demand procedures, message channels (HTTP, Slack, Discord) and recurring cron jobs. Optional agent.ts sets the model and reasoning level.
The framework is set up through the npx eve init command, which scaffolds a new project, installs dependencies and starts an interactive terminal UI. Tools are defined with zod input schemas, and the default model is routed through the Vercel AI Gateway. Web Chat can be added with npx eve add channel/web, with sign-in through Vercel for team members and deployment via npx eve deploy.
The project is written in TypeScript, licensed under Apache-2.0, and is currently in beta, so APIs and behavior may change before general availability.
Key features
- 01Defines agent instructions, tools, skills, channels and schedules as files in a fixed directory layout.
- 02Provides typed tools with zod input schemas that the model can call.
- 03Supports HTTP, Slack and Discord message channels.
- 04Runs recurring cron jobs through schedules/ files.
- 05Includes a web chat channel with optional Vercel sign-in and a deploy command.
Use cases
- Building a custom AI assistant that calls your own functions and APIs.
- Deploying an agent that answers messages in Slack or Discord.
- Running scheduled agent jobs such as weekly summaries.
- Creating a team-accessible web chat front end for an internal agent.
Quick start
- 1
Create a new agent project with the interactive setup
npx eve@latest init my-agent - 2
Start the agent locally after replacing the instructions and tool
npm run dev
Things to know
- The project is in beta and its APIs and behavior may change before general availability.
- Default setup uses an AI Gateway model, so you need model access to run the agent.
Topics
FAQ
- What is eve?
- A filesystem-first TypeScript framework for building durable AI agents, with tools, skills, channels and schedules defined as files. eve is a framework for building AI agents where the project structure is the authoring interface. An agent lives in an agent/ directory: instructions.md holds the always-on system prompt, while tools/, skills/, channels/ and schedules/ hold typed tool functions, on-demand procedures, message channels (HTTP, Slack, Discord) and recurring cron jobs. Optional agent.ts sets the model and reasoning level.
- Is eve free to use? What license does it have?
- eve is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
- Is eve still maintained?
- Actively maintained. The last commit to eve was on Oct 22, 2026.
- What can I use eve for?
- Building a custom AI assistant that calls your own functions and APIs. Deploying an agent that answers messages in Slack or Discord. Running scheduled agent jobs such as weekly summaries. Creating a team-accessible web chat front end for an internal agent.
- How do I install eve?
- Create a new agent project with the interactive setup: npx eve@latest init my-agent Start the agent locally after replacing the instructions and tool: npm run dev
- Where can I get eve?
- The source code and installation instructions are on GitHub: https://github.com/vercel/eve
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
Open-source infrastructure for running AI-generated code in isolated cloud sandboxes, with JavaScript and Python SDKs.
A TypeScript framework for building AI agents and LLM-powered applications, with workflows, tools, memory, MCP servers, and evals.
Open-source long-horizon super agent harness that orchestrates sub-agents, memory, and sandboxes to research, code, and create across tasks lasting minutes to hours.
A minimal AI software engineering agent that solves GitHub issues or helps in the command line, using only bash and about 100 lines of core Python.
Open-source, self-hostable AI operating system where agents run in isolated cloud sandboxes and deliver work through approved change requests.